We show the powerful sparse signal recovery approach FOCUSS is a convex-concave procedure. It follows a Newton-like decent direction by retaining the positive definite component of the Hessian matrix. This motivates an improved implementation of FOCUSS.